To test investment management skill, the authors compare the performance of professionally managed long-only US equity funds with portfolios randomly generated from the S&P 500 Index universe. The methodology they use produces findings that indicate that the professionally managed portfolios do not have a higher probability of outperforming the index than the randomly generated portfolios do.
Using an approach similar to Roy’s safety-first risk rule
(Econometrica 1952), the authors randomly generate portfolios using the
S&P 500 Index as the universe of available stocks. To test investment management skill,
the randomly generated portfolios are compared with a large sample of professionally managed
portfolios consisting of US long-only large-cap stocks. Although the results are consistent
with those of previous findings (i.e., that markets are efficient), the authors use an
innovative approach to conducting their experiment. They use a probability-based analysis
and explore the implications of such an approach.
This research is useful to practitioners who invest in long-only equity portfolios with a
turnover ratio of around 100%. The turnover ratio of 100% is represented by one of the
assumptions in the analysis, which stipulates that each randomly generated portfolio be sold
at each year-end and then randomly repurchased at the beginning of the year.
For the universe and investment horizon tested, the results indicate that a naive beta
strategy (i.e., a randomly generated portfolio) is probabilistically geared to perform as
well as a consciously managed portfolio. Some outperformance of the markets is expected
because the portfolios are not risk adjusted. Neither the random portfolios nor the managed
portfolios follow a particular pattern of performance. The more concentrated random
portfolios tend to outperform in years other than times of crisis or bear market rallies.
The tech bubble benefited managed portfolios and weakened the results of the more
diversified random portfolios. The authors suggest that asset managers may lack tools that
have predictive power or may be constrained by mandates, although they do tend to outperform
during bull markets.
The focus of previous studies of actively managed portfolios has generally been on the
efficiency of the markets (comparing actively managed portfolios with passive portfolios).
To reflect strategies used in extreme times, a more recent study compared three portfolio
strategies (random, index benchmark, and smart beta) with the markets by using the actual
returns of professional managers. In addition, an attempt has been made to measure the
determinants of skill or luck. The authors of this research consider whether it is either
skill or luck that determines performance.
For the investment period from 1 January 1991 to the end of 2011, the authors randomly
select stocks from each year to construct their test portfolios. To avoid selection bias,
the authors use Mercer’s Global Investment Manager Database (GIMD) in choosing the
actively managed accounts. A sample size of 10,000 randomly generated portfolios is chosen
to allow for a more accurate inference of probability density functions while avoiding
selection bias. The proportion of each stock in the portfolio is also random but must be
greater than 0% and less than 100%. A portfolio of 30 stocks is chosen because the marginal
benefits of diversification are found to become nil beyond 30; a portfolio of 3 stocks is
chosen to demonstrate the idiosyncratic effects, and a portfolio of 100 stocks is chosen to
demonstrate the full benefits of diversification while being less than the investment
universe of 500 stocks.
To follow the reality of the market over the 20-year time period of the study, the
portfolios are turned over annually. To study the effects of various economic cycles, the
authors use the Akaike information criterion to assign cross-sectional probability density
functions that best describe the performance of both random and managed portfolios. They
state that a significant number of the inferred distributions are nonnormal.
The authors conclude that randomly selected portfolios perform just as well as actively
managed portfolios. But these results are limited by the assumptions they make. The authors
do note that the results cannot necessarily be extrapolated to asset managers in general and
are only representative of the asset managers in the sample. It is unclear whether the
assumptions used to create the random portfolios fairly represent the characteristics
inherent to managed portfolios, which may have different holding periods, and whether the
universe of managed portfolios is limited to that of the S&P 500.